CN101763527B - Method for detecting number based on side texture analysis of layered object - Google Patents
Method for detecting number based on side texture analysis of layered object Download PDFInfo
- Publication number
- CN101763527B CN101763527B CN 200910087353 CN200910087353A CN101763527B CN 101763527 B CN101763527 B CN 101763527B CN 200910087353 CN200910087353 CN 200910087353 CN 200910087353 A CN200910087353 A CN 200910087353A CN 101763527 B CN101763527 B CN 101763527B
- Authority
- CN
- China
- Prior art keywords
- image
- texture
- layered object
- sub
- images
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000004458 analytical method Methods 0.000 title abstract description 7
- 238000001514 detection method Methods 0.000 claims abstract description 22
- 238000012545 processing Methods 0.000 claims abstract description 20
- 238000003709 image segmentation Methods 0.000 claims abstract description 13
- 238000005070 sampling Methods 0.000 claims abstract description 10
- 230000002708 enhancing effect Effects 0.000 claims abstract description 8
- 238000001914 filtration Methods 0.000 claims description 24
- 230000008569 process Effects 0.000 claims description 11
- 230000000737 periodic effect Effects 0.000 claims description 8
- 230000011218 segmentation Effects 0.000 claims description 7
- 238000005520 cutting process Methods 0.000 claims description 5
- 238000013459 approach Methods 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 4
- 125000002015 acyclic group Chemical group 0.000 claims description 3
- 241000282341 Mustela putorius furo Species 0.000 claims description 2
- 238000005192 partition Methods 0.000 claims description 2
- 230000001965 increasing effect Effects 0.000 abstract description 4
- 238000004519 manufacturing process Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 238000005728 strengthening Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 238000003754 machining Methods 0.000 description 1
- 230000035800 maturation Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a method for detecting number based on side texture analysis of a layered object. The method comprises the following processing steps: side texture of an object to be detected is sampled and then after sampling, the image is input to a computer; the original image is pre-processed and sectioned, thus obtaining a sub-image set; image segmentation is carried out on a section of sub-images in the sub-image set to obtain a corresponding foreground texture image; the foreground texture image is projected or sampled to obtain a one-dimensional array; the one-dimensional array is analyzed to obtain an estimated number value of the sub-images, and the estimated value is stored to be used; all sub-images in the sub-image set is circularly processed; and the estimated number value of the sub-images is counted to obtain the number of the layered object. The method introduces texture analysis into number detection of the layered object and applies a Gabor filter skillfully to carry out texture enhancement, thus enhancing performance of the algorithm. The sectioning processing divides the original texture image into multiple sections of sub-images, thus increasing image segmentation precision and enhancing detection accuracy.
Description
Technical field
The present invention relates to image processing field, design and realized a kind of number detection method of analyzing based on zonal texture.
Background technology
In production field, it is the important ring that product quality detects that quantity detects.Particularly the number of layered object detects, and is the problem that lets the people perplex always, exists cost and efficient contradiction because number detects.This contradiction mainly shows following three aspects: the first, artificial mode goes the number of the product of every day 1,000,000 is verified it is infeasible one by one; So often adopt the method for sampling Detection in the practical application; This will certainly cause omission, has increased the risk that substandard products occur.Even if second sampling Detection, in the face of a large amount of product like this, it also is quite high detecting needed human resources and handling cost.Three, this type of work is that high-intensity repeated labor is easy to make the people to produce fatigue, thereby produces the flase drop of product, has increased detection risk.
Developing rapidly of, hardware technology soft along with computing machine, the maturation of image processing techniques, and the cost performance of industrial camera improves, machine vision technique emerges rapidly.Machine vision utilizes image processing techniques to combine artificial intelligence, comes the vision of simulating human, measures and judgement thereby utilize machine to replace human eye to do.These technological characteristics are to improve the flexibility and the automaticity of producing.Not too be suitable for the occasion that manual work working environment or artificial vision are difficult to meet the demands at some, the machine in normal service vision substitutes the artificial vision; Simultaneously in industrial processes in enormous quantities, low and precision is not high with artificial visual inspection product quality efficient, the automaticity that can enhance productivity greatly and produce with machine vision detection method.And machine vision is easy to realize that information integration is the basic technology that realizes computer integrated manufacturing system.Can obtain bulk information fast just because of NI Vision Builder for Automated Inspection; And be easy to automatic processing; Also be easy to design information and machining control information integration; Therefore, in the automated production process in modern times, people are widely used for NI Vision Builder for Automated Inspection in fields such as operating condition monitoring, product inspection and quality control.
Summary of the invention
The objective of the invention is to utilize image processing techniques through the side grain analysis of layered object being obtained the number of the target that detects, a kind of number detection method of analyzing based on the layered object side grain is provided for this reason.
In order to reach said purpose, the present invention provides the number detection method of analyzing based on the layered object side grain, and the step of this method is following:
Step 1: gather the side grain image of layered object to be detected by industrial camera or digitized instrument, and the side grain image is converted into digital picture; Through interface unit digital picture is read in the image processing system, adopts the Gabor wave filter to carry out Filtering Processing to digital picture then, generate texture and strengthen image, again texture enhancing image is carried out staging treating and obtain set of sub-images;
Step 2: get the cross-talk image in the set of sub-images after the segmentation, the utilization image partition method is partitioned into the texture region that will detect, and generates corresponding target texture image;
Step 3: from the target texture image, extract the one-dimension array that characterizes the number texture;
Step 4: the above-mentioned one-dimension array of analyzing and processing, obtain the number estimated value of target texture image, preserve this estimated value;
Step 5: whether all subimages in the set of sub-images of determining step 2 dispose, if dispose, then execution in step 6, if do not handle, then return step 2 circular treatment;
Step 6: all estimated values of preserving in the statistic procedure 4 obtain the number of target.
Wherein, the side grain image of this layered object is comprising and has periodically variable zonal texture; Texture in the side grain image of this layered object has the direction consistance; Have acyclic grey scale change or interference noise on the side grain image of this layered object.
Wherein, set of sub-images is that texture enhancing image is to obtain along the direction segmentation that texture extends.
Wherein, image segmentation adopts the dividing method based on the edge, promptly is the border of coming ferret out texture image and background image through the straight line in the match edge, through boundary definition target texture zone.
Wherein, the one-dimension array that extracted of target texture image has following characteristic:
1. this array is the one dimension projection of texture information or along the sampling of certain direction;
2. the one dimension projecting direction is along the texture bearing of trend, and image sampling is then along the direction of texture periodic extension.
Wherein, the analyzing and processing process to one-dimension array comprises the steps:
1. to one-dimension array frequency domain or spatial domain carry out low-pass filtering treatment obtain filtering after array;
2. crest in the array after the filtering or trough are counted, obtained the number estimated value.
Wherein, the number of target is to ask mode to obtain to all estimated values.
Beneficial effect of the present invention: ultimate principle of the present invention is to utilize the number of times of individual texture appearance of repetition period property calculating of layered object side grain.The invention property ground is incorporated into texture analysis in the number detection of layered object, forms the complete frame of a disposable detection number.The present invention has carried out the texture enhancing with the Gabor wave filter to original image, thereby has improved the performance of algorithm.The present invention is with the 2 d texture projection or be sampled as one-dimension array, under the prerequisite that guarantees analytical effect, has improved detection speed greatly.The present invention is incorporated into LPF in the number counting dexterously, has successfully solved the number that has than the strong jamming image and has detected problem.The introducing that the present invention is successful the thought of staging treating, former texture image is divided into many cross-talk images, both improved the image segmentation precision, strengthened the accuracy that detects again.Algorithm application provided by the present invention is in extensive range, is suitable for the various periodically variable zonal texture enumeration problems that have.
Description of drawings
Fig. 1 is the process flow diagram of technical scheme of the present invention;
Fig. 2 a-Fig. 2 d is effect contrast figure and an image segmentation synoptic diagram before and after the Gabor filtering;
Fig. 3 a-Fig. 3 c is the image segmentation process of certain cross-talk image;
Fig. 4 a-4b is the acquisition process of one-dimension array;
Fig. 5 is the filtered curve map of one-dimension array;
Fig. 6 is pack side grain figure.
Embodiment
Specify each related detailed problem in the technical scheme of the present invention below in conjunction with accompanying drawing.Be to be noted that described instance only is intended to be convenient to understanding of the present invention, and it is not played any qualification effect.
Ultimate principle of the present invention is to utilize the number of times of individual texture appearance of repetition period property calculating of layered object side grain, detects the process flow diagram that step sees also Fig. 1 technical scheme of the present invention.
The present invention gathers the side image of layered object to be detected with industrial camera or other digitized instruments, and is translated into digital picture; Again digital picture is imported in the image processing system through USB interface; Carry out number at last and detect and export the result.The invention is characterized in and comprise the steps:
(1) image capturing system is gathered target image to be detected, and the side grain image that will detect target image is transported in the image processing system through interfaces such as USB.This texture image has following characteristic:
1. Fig. 2 a is the side grain of stratiform measured target image, is comprising periodic number information in this texture;
2. possibly there are various acyclic grey scale change and some random noises on this side grain image;
3. the thickness of each layer is about the same in the stratiform measured target image, every layer of about several to dozens of pixel;
4. each layer is compact arranged in the side grain image, and texture has very strong directivity and periodicity;
5. target side texture image has the bearing of trend of texture periodic extension direction and texture, and each layer is a primitive in the texture image, and the direction of all primitive periodic arrangement is the periodic extension direction, and the direction that single primitive extends is the bearing of trend of texture.
(2) original image shown in Fig. 2 a is carried out the texture enhancement process, this process is to adopt the Gabor wave filter to carry out filtering.The direction of Gabor wave filter is identical with the direction that texture extends, and supposes that each layer width of stratiform target image is approximately n pixel, and then the width of Gabor wave filter also is n.The size of window is W*W, wherein W=2n+1 during filtering.The frequency response chart of Gabor wave filter is seen Fig. 2 b.With the wave filter that we constructed image is carried out filtering and obtain image such as Fig. 2 c after texture strengthens.Then, the image after strengthening is carried out segmentation, the direction of segmentation is consistent with texture periodic extension direction.Adopt transversal sectional to obtain the set of subimage to strengthening texture image in this example, shown in Fig. 2 d.Image segmentation is handled has following advantage:
1. edge of image is more near straight line after the segmentation, and direction fixes, thereby helps increasing through edge fitting the accuracy of image segmentation.
2. behind the image segmentation, can carry out repeatedly number to texture image and estimate, confirm final result through the statistics mode at last.Help avoiding the single number to estimate the accidental error that occurs like this, make result's robust more.
(3) in this step each subimage among Fig. 2 d being carried out circular treatment disposes up to all subimages and just gets into next step.Processing procedure comprises that image segmentation, image projection, drop shadow curve handle, number detects.In order to specify above-mentioned processing procedure, we get among Fig. 2 d any one section image as an example.Adopt the mode that the edge is cut apart to be partitioned into the texture region that will detect among the present invention.If cut apart before certain number of sub images be I (x, y) shown in Fig. 3 a, x, y are the coordinate figure of pixel in the image, the I representative image is at the gray-scale value of this point.It is used the Canny operator, and the detected image edge obtains corresponding edge image E, and (x is y) like Fig. 3 b.We can see between texture and black background from outline map has two vertical short lines, and we are referred to as the border.Through the search border target texture image in the image is defined out, cut apart back target texture image shown in Fig. 3 c.
Ensuing work is to extract the one-dimension array that can characterize the number texture the foreground image Fig. 3 c after cutting apart, and Fig. 4 a-Fig. 4 b has showed the process of obtaining this one-dimension array.This array obtained two kinds of approach:
1. Fig. 3 c is carried out projection, the direction of image projection is identical with the direction that texture extends, and Fig. 4 a is the one-dimension array that projection obtains.
2. Fig. 3 c is sampled, the sampling direction of curve is along the periodic extension direction of texture, and Fig. 4 b is the one-dimension array that sampling obtains.
Wherein, the one-dimension array that obtains contains all number texture information projections, and the fluctuating of the Wave crest and wave trough of this one-dimension array has highlighted the cyclical variation of number texture.
Through analyzing the one-dimension array (what analyze in this example is drop shadow curve) that projection or sampling obtain, obtain the corresponding number estimated value of subimage Fig. 3 c then.With reference to the concrete narration of the above-mentioned analytic process of Fig. 5 as follows:
1. the dimension curve shown in Fig. 4 a is carried out LPF:
Because the irrelevant grey scale change (like Fig. 2) of noise or original image itself has a lot of interference components on the resulting dimension curve.This produces error when carrying out crest or trough counting, thereby causes faults.Therefore, curve needs low-pass filtering treatment before counting.Filtering can be adopted frequency filtering, also can carry out filtering in spatial domain.The present invention's employing [1/n, 1/n, 1/n ... 1/n, 1/n, 1/n, 1/n] be the filtering template, wherein n is the length of filter window, spatial domain and curve carry out convolution obtained filtering after the curve (see figure 5).
2. curve gets number after the analysis filtered:
Number to crest in the curve after the filtering or trough is counted, and obtains the number value at last.In the middle of practical application, still be that trough is counted and need be decided as the case may be to crest.One-dimension array medium wave peak valley detection obtains through array after the filtering is carried out second order difference.
(4) add up the many corresponding number estimated values of all subimages, confirm the final number of target through asking mode (the highest value of the frequency of occurrences in all estimated values).
Based on algorithm provided by the present invention, we have done a playing card number detection system, and this system uses industrial camera, this camera collection be 24 gray-scale maps, size is 1280*1024 (as shown in Figure 6).It is to be detected to import computing machine etc. into through USB interface after the images acquired.We analyze for ease and improve arithmetic speed, and pending image has been done following three hypothesis:
The correct number of the playing card that 1) detected is fixed as 55 (54 add billboard);
2) fixed angle in playing card side when imaging and be approximately 90 °;
3) target and background discrimination to be detected is high in the image of being gathered, and background is clean, does not have other interference of texture.
Implementation step once is described below:
The first step: original image is carried out pre-service, utilize the Gabor wave filter to carry out texture here and strengthen.According to top hypothesis, the direction of Gabor wave filter is 90 °, and its fixed width is 8 pixels, is about the thickness of a sheet playing card.Window size is a W=2*8+1=17 pixel.After the enhancing image is cut into 32 cross-talk images, every section size is 1280*32.
Next circulate above-mentioned subimage is handled, obtain the corresponding number estimated value of each subgraph.This processing procedure comprises: image segmentation, image projection, drop shadow curve's filtering, drop shadow curve's Wave crest and wave trough counting.Explain that for ease we are split as following step 2 to this process to step 4.
Second step: a cross-talk image of getting in the set of sub-images carries out image segmentation; Cutting procedure adopts the edge search; The vertical short-term (seeing the pink lines of Fig. 3 b) that promptly occurs for the first time from image left and right-hand search is respectively regarded them the border, the left and right sides of texture as.
The 3rd step: the image after will cutting apart is to the horizontal direction projection, and (Fig. 4 a) to obtain one-dimension array X (n).
The 4th step: X (n) is carried out low-pass filtering treatment, and we are directly in spatial domain filtering here, and wave filter is [1/8,1/8,1/8,1/8,1/8,1/8,1/8,1/8 ,] obtain curve (Fig. 4 b) after the filtering, then the curve trough is counted to get number estimated value N.
The 5th step: judge whether all subimages in the set of sub-images dispose, then got into for the 6th step, otherwise return the second step circular treatment if dispose.
The 6th step: add up all estimated values, obtain the number of target, export result, detection of end at last.
The above; Be merely the embodiment among the present invention, but protection scope of the present invention is not limited thereto, anyly is familiar with this technological people in the technical scope that the present invention disclosed; Can understand conversion or the replacement expected; All should be encompassed in of the present invention comprising within the scope, therefore, protection scope of the present invention should be as the criterion with the protection domain of claims.
Claims (6)
1. number detection method of analyzing based on the layered object side grain is characterized in that:
Step 1: gather the side grain image of layered object to be detected by industrial camera or digitized instrument, and the side grain image is converted into digital picture; Through interface unit digital picture is read in the image processing system, adopts the Gabor wave filter to carry out Filtering Processing to digital picture then, generate texture and strengthen image, again texture enhancing image is carried out staging treating and obtain set of sub-images;
Step 2: get the cross-talk image in the set of sub-images after the segmentation, the utilization image partition method is partitioned into the texture region that will detect, and generates corresponding target texture image;
Step 3: from the target texture image, extract the one-dimension array that characterizes the number texture; One-dimension array is obtained following two kinds of approach: approach one is that the foreground image after cutting apart is carried out projection, and the direction of image projection is identical with the direction that texture extends, the one-dimension array that projection obtains; Approach two is that the foreground image after cutting apart is sampled, and the sampling direction of curve is along the periodic extension direction of texture, the one-dimension array that sampling obtains;
Step 4: the above-mentioned one-dimension array of analyzing and processing, obtain the number estimated value of target texture image, preserve this estimated value;
Step 5: whether all subimages in the set of sub-images of determining step 2 dispose, if dispose, then execution in step 6, if do not handle, then return step 2 circular treatment;
Step 6: all estimated values of preserving in the statistic procedure 4 obtain the number of target.
2. the number detection method of analyzing based on the layered object side grain according to claim 1 is characterized in that, the side grain image of this layered object is comprising and has periodically variable zonal texture; Texture in the side grain image of this layered object has the direction consistance; Have acyclic grey scale change or interference noise on the side grain image of this layered object.
3. the number detection method of analyzing based on the layered object side grain according to claim 1 is characterized in that, set of sub-images is that texture enhancing image obtains along the direction segmentation that texture extends.
4. the number detection method of analyzing based on the layered object side grain according to claim 1; It is characterized in that; Image segmentation adopts the dividing method based on the edge; Promptly be the border of coming ferret out texture image and background image through the straight line in the match edge, through boundary definition target texture zone.
5. the number detection method of analyzing based on the layered object side grain according to claim 1 is characterized in that, the analyzing and processing process of one-dimension array is comprised the steps:
1. to one-dimension array frequency domain or spatial domain carry out low-pass filtering treatment obtain filtering after array;
2. crest in the array after the filtering or trough are counted, obtained the number estimated value.
6. the number detection method of analyzing based on the layered object side grain according to claim 1 is characterized in that the number of target is to ask mode to obtain to all estimated values.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200910087353 CN101763527B (en) | 2009-06-17 | 2009-06-17 | Method for detecting number based on side texture analysis of layered object |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200910087353 CN101763527B (en) | 2009-06-17 | 2009-06-17 | Method for detecting number based on side texture analysis of layered object |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101763527A CN101763527A (en) | 2010-06-30 |
| CN101763527B true CN101763527B (en) | 2012-12-12 |
Family
ID=42494683
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200910087353 Expired - Fee Related CN101763527B (en) | 2009-06-17 | 2009-06-17 | Method for detecting number based on side texture analysis of layered object |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101763527B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1607551A (en) * | 2003-08-29 | 2005-04-20 | 三星电子株式会社 | Method and apparatus for image-based photorealistic 3D face modeling |
| CN1728161A (en) * | 2005-07-28 | 2006-02-01 | 上海交通大学 | Method for filtering sensing images based on heteropic quantized color feature vectors |
-
2009
- 2009-06-17 CN CN 200910087353 patent/CN101763527B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1607551A (en) * | 2003-08-29 | 2005-04-20 | 三星电子株式会社 | Method and apparatus for image-based photorealistic 3D face modeling |
| CN1728161A (en) * | 2005-07-28 | 2006-02-01 | 上海交通大学 | Method for filtering sensing images based on heteropic quantized color feature vectors |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101763527A (en) | 2010-06-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109087274B (en) | Electronic device defect detection method and device based on multi-dimensional fusion and semantic segmentation | |
| CN108154105B (en) | Underwater biological detection and identification method and device, server and terminal equipment | |
| CN109447945B (en) | Quick counting method for basic wheat seedlings based on machine vision and graphic processing | |
| CN108230321A (en) | Defect inspection method and device | |
| Toh et al. | Automated fish counting using image processing | |
| CN109507192A (en) | A kind of magnetic core detection method of surface flaw based on machine vision | |
| CN107220649A (en) | A kind of plain color cloth defects detection and sorting technique | |
| WO2010063252A1 (en) | Method for measuring the growth of leaf disks of plants and apparatus suited therefor | |
| CN115330784A (en) | Cloth surface defect detection method | |
| CN113610772B (en) | Method, system, device and storage medium for detecting spraying code defect at bottom of pop can bottle | |
| CN110717909A (en) | Metal surface scratch detection method and device | |
| CN111353992B (en) | Agricultural product defect detection method and system based on textural features | |
| Maheshwari et al. | Non-destructive quality analysis of Indian Gujarat-17 Oryza sativa SSP Indica (Rice) using image processing | |
| CN102496032B (en) | Electrical equipment X ray digital image processing algorithm support system | |
| CN111221996B (en) | Instrument screen vision detection method and system | |
| Mamatov et al. | Algorithm for extracting contours of agricultural crops images | |
| CN108107611A (en) | A kind of adaptive defect inspection method, device and electronic equipment | |
| EP3510526B1 (en) | Particle boundary identification | |
| Maheshwari et al. | Non-destructive quality analysis of Indian Basmati Oryza sativa SSP indica (Rice) using image processing | |
| CN101763527B (en) | Method for detecting number based on side texture analysis of layered object | |
| CN109544514A (en) | A kind of sawn timber identity identification method, device and equipment merging performance characteristic | |
| Maheshwari | Quality assessment of oryza sativa ssp indica (rice) using computer vision | |
| Yang et al. | Cherry recognition based on color channel transform | |
| Niskanen et al. | Experiments with SOM based inspection of wood | |
| JP7460391B2 (en) | LEARNING DEVICE, IMAGE INSPECTION DEVICE, PROGRAM, AND LEARNING METHOD |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20121212 Termination date: 20210617 |